Pulse24 Original
Half of Wall Street Called AI a Risk in May. By October, 87% Did.
November 14, 2026
The Fed's twice-yearly check-in with market contacts just produced its sharpest swing in years: AI jumped from a risk half of them mentioned in May to one 87% of them cite now, alongside a fivefold rise in hyperscaler debt since 2023. The central bank flags the exposure behind the AI buildout as bigger than it looks on paper, well short of forecasting a crash.
Twenty-three people. That's the entire panel the Federal Reserve quietly polls twice a year to find out what might break the financial system next. Their answer this fall, tucked into the central bank's November 2026 Financial Stability Report released Friday, was about as close to unanimous as that survey gets. Eighty-seven percent of those contacts flagged artificial intelligence as a salient risk to financial stability. Back in May, only half of them did.
[[IMG1]]
A Fivefold Jump Since 2023
The survey swing is the headline, but the Fed backed it with numbers that explain why risk officers are suddenly so uneasy. Debt issuance tied to AI data centers has risen to more than $680 billion as of October, the report says, up from roughly $130 billion across all of 2023. That's not a gradual climb. It's a market that didn't meaningfully exist three years ago now rivaling entire industries for credit market attention.
What worries the Fed more than the size of that number is its shape. Financing has drifted away from plain corporate bonds, the kind large, cash-generative companies have issued for decades, toward a mix of public debt, private credit, and structured deals. Those structures increasingly link hyperscalers and GPU designers to smaller firms with thinner balance sheets and less history in credit markets. CoreWeave has become the industry's go-to example of what that financing looks like up close: a company whose earnings calls now spend as much time defending its debt costs to bondholders as explaining growth to analysts.
On top of the debt itself, the Fed estimates future lease commitments and backstop guarantees tied to AI build-outs now total about $3 trillion. Leases and guarantees don't show up on a balance sheet the way a bond does, but they still represent real obligations. If the AI investment cycle slows, that gap between what's visible on paper and what's actually owed is where the report says the damage could concentrate.
Why This Matters for Credit Markets
None of this means the Fed is predicting a crash. The report reads more like an insurance adjuster flagging a policy that's been underpriced than a forecaster calling a top. Still, the mechanics matter for anyone watching credit spreads. The widening in high-yield debt earlier this month, traced largely to a single Texas data center bond, was a small preview of what the Fed is describing at scale: a market where one weak link in the AI financing chain moves pricing well beyond that single issuer.
Hedge funds add another layer. The report notes they're carrying their highest leverage since 2013, much of it built on the same assumption underpinning the whole AI trade: that demand for compute keeps growing fast enough to service the debt raised to build it. Banks that sell off data center loan exposure, the way lenders have increasingly done through structured auctions and portfolio sales, are partly managing that same concentration risk rather than creating new ones from scratch.
Equity investors have been pricing in some version of this unease already. The Nasdaq-100's repeated pullbacks this year, including its third correction this year on doubts about AI capital spending, reflect a market that keeps testing how much AI-linked risk it's willing to hold at once. The Fed's survey suggests the credit side of that same trade has been quietly catching up to the stock market's nerves.
What to Watch Next
Three things will say more about whether this risk stays contained or spreads. First, whether the pace of new AI-linked debt issuance slows from its current run rate, or whether lenders keep extending credit to smaller, less-established firms in the supply chain. Second, whether credit spreads on data center and hyperscaler-linked debt widen further from here, which would suggest the market is already repricing the risk the Fed just described. Third, whether hedge fund leverage comes down from its highest level since 2013 on its own, or whether it takes a sharper move in AI-related assets to force the unwind.
The Fed's own framing is useful here: it isn't treating AI spending as a bubble that must pop, but as a leverage buildup whose true size is harder to see than usual. That's a different kind of risk than a simple valuation call, and it's one that shows up first in funding markets and loan pricing before it ever reaches a headline stock chart.
The Pulse24 Take
Strategists have questioned AI valuations for two years already. The notable part of this report is the Fed putting a number on how fast that unease has spread among the people who actually price credit risk for a living, attaching real figures, the $680 billion in debt and the $3 trillion in contingent commitments, to what had mostly stayed a qualitative worry until now.
The sensible takeaway for investors isn't a call to assume the AI buildout is about to unwind; the financing behind it has simply grown more complex and more opaque at the same time it's grown larger, and that kind of complexity is exactly what regulators tend to flag before markets fully price it. Watching credit spreads on data center debt from here on may say more about the health of the AI trade than any single company's stock chart.
How we read the data
Curious how we get from raw data to a take like this? Our Trader's Toolkit walks through the tools we lean on.
Explore the Toolkit